Data Visualization with ggplot2

ggplot2 is based on the grammar of graphics, the idea that you can build every graph from the same few components: a data set, a set of geoms—visual marks that represent data points, and a coordinate system

Package and imported data

# install.packages('ggplot2')
library(ggplot2)

library(readr)
nba_position <- read_csv("~/Desktop/Microsoft/DataVisualization/nba_player_position.csv")
nba_players <- read_csv("~/Desktop/Microsoft/DataVisualization/nba_raw_players.csv")
nba_id_players <- read_csv("~/Desktop/Microsoft/DataVisualization/nba_id_players.csv")

Elements of graph

Layer(層)

Data：使用什麼資料
Mapping：使用資料裡什麼變數
Statistical transformation (stat)：需不需要對變數先做統計轉換
Geometric object (geom)：要畫什麼幾何圖形（點、線等等）
Position adjustment (position)：幾何圖形位置調整

Scale(刻度尺)

Coordinate system (座標系統)

Faceting (多圖面)

canvas of ggplot

ggplot()會製造出一張空白畫布，供我們疊加更多圖層。

ggplot(nba_position, aes(HEIGHT, WEIGHT))

Data Preprocessing

nba_position$POSITION[nba_position$POSITION == "Forward-Center"] <- "Center-Forward"
nba_position$POSITION[nba_position$POSITION == "Guard-Forward"] <- "Forward-Guard"

選出 LeBron James 的生涯數據資料：

library(dplyr)
LeBron_ID <- nba_id_players[which(nba_id_players$Name=='LeBron James'),]$id
LeBron <- filter(nba_players, PLAYER_ID==LeBron_ID)
show(LeBron)

## # A tibble: 16 x 27
##    PLAYER_ID SEASON_ID LEAGUE_ID TEAM_ID TEAM_ABBREVIATI… PLAYER_AGE    GP    GS
##        <dbl> <chr>         <dbl>   <dbl> <chr>                 <dbl> <dbl> <dbl>
##  1      2544 2003-04           0  1.61e9 CLE                      19    79    79
##  2      2544 2004-05           0  1.61e9 CLE                      20    80    80
##  3      2544 2005-06           0  1.61e9 CLE                      21    79    79
##  4      2544 2006-07           0  1.61e9 CLE                      22    78    78
##  5      2544 2007-08           0  1.61e9 CLE                      23    75    74
##  6      2544 2008-09           0  1.61e9 CLE                      24    81    81
##  7      2544 2009-10           0  1.61e9 CLE                      25    76    76
##  8      2544 2010-11           0  1.61e9 MIA                      26    79    79
##  9      2544 2011-12           0  1.61e9 MIA                      27    62    62
## 10      2544 2012-13           0  1.61e9 MIA                      28    76    76
## 11      2544 2013-14           0  1.61e9 MIA                      29    77    77
## 12      2544 2014-15           0  1.61e9 CLE                      30    69    69
## 13      2544 2015-16           0  1.61e9 CLE                      31    76    76
## 14      2544 2016-17           0  1.61e9 CLE                      32    74    74
## 15      2544 2017-18           0  1.61e9 CLE                      33    82    82
## 16      2544 2018-19           0  1.61e9 LAL                      34    55    55
## # … with 19 more variables: MIN <dbl>, FGM <dbl>, FGA <dbl>, FG_PCT <dbl>,
## #   FG3M <dbl>, FG3A <dbl>, FG3_PCT <dbl>, FTM <dbl>, FTA <dbl>, FT_PCT <dbl>,
## #   OREB <dbl>, DREB <dbl>, REB <dbl>, AST <dbl>, STL <dbl>, BLK <dbl>,
## #   TOV <dbl>, PF <dbl>, PTS <dbl>

選出 Dwyane Wade 的生涯數據資料：

Wade_ID <- nba_id_players[which(nba_id_players$Name=='Dwyane Wade'),]$id
Dwyane <- filter(nba_players, PLAYER_ID==Wade_ID & TEAM_ABBREVIATION != 'TOT')
show(Dwyane)

## # A tibble: 17 x 27
##    PLAYER_ID SEASON_ID LEAGUE_ID TEAM_ID TEAM_ABBREVIATI… PLAYER_AGE    GP    GS
##        <dbl> <chr>         <dbl>   <dbl> <chr>                 <dbl> <dbl> <dbl>
##  1      2548 2003-04           0  1.61e9 MIA                      22    61    56
##  2      2548 2004-05           0  1.61e9 MIA                      23    77    77
##  3      2548 2005-06           0  1.61e9 MIA                      24    75    75
##  4      2548 2006-07           0  1.61e9 MIA                      25    51    50
##  5      2548 2007-08           0  1.61e9 MIA                      26    51    49
##  6      2548 2008-09           0  1.61e9 MIA                      27    79    79
##  7      2548 2009-10           0  1.61e9 MIA                      28    77    77
##  8      2548 2010-11           0  1.61e9 MIA                      29    76    76
##  9      2548 2011-12           0  1.61e9 MIA                      30    49    49
## 10      2548 2012-13           0  1.61e9 MIA                      31    69    69
## 11      2548 2013-14           0  1.61e9 MIA                      32    54    53
## 12      2548 2014-15           0  1.61e9 MIA                      33    62    62
## 13      2548 2015-16           0  1.61e9 MIA                      34    74    73
## 14      2548 2016-17           0  1.61e9 CHI                      35    60    59
## 15      2548 2017-18           0  1.61e9 CLE                      36    46     3
## 16      2548 2017-18           0  1.61e9 MIA                      36    21     0
## 17      2548 2018-19           0  1.61e9 MIA                      37    72     2
## # … with 19 more variables: MIN <dbl>, FGM <dbl>, FGA <dbl>, FG_PCT <dbl>,
## #   FG3M <dbl>, FG3A <dbl>, FG3_PCT <dbl>, FTM <dbl>, FTA <dbl>, FT_PCT <dbl>,
## #   OREB <dbl>, DREB <dbl>, REB <dbl>, AST <dbl>, STL <dbl>, BLK <dbl>,
## #   TOV <dbl>, PF <dbl>, PTS <dbl>

將賽季切分成兩個年份，以賽季開始年份為主：

library(tidyr)
LeBron <- LeBron %>% separate(SEASON_ID, 
                              c("YEAR", "REST"))
Dwyane <- Dwyane %>% separate(SEASON_ID, 
                              c("YEAR", "REST"))

Aesthetics

ggplot2 當中用來美化視覺化結果的 mapping，例如設定顏色、形狀、透明度、大小等等，更多用法可見 Aesthetic specifications.

ggplot(nba_position, aes(HEIGHT, WEIGHT))+
  geom_point()

Color, Alpha, Shape, …

ggplot(nba_position, aes(HEIGHT, 
                         WEIGHT, 
                         color = POSITION, 
                         alpha = 0.3,
                         shape = POSITION
                         # size = ...
                         ))+
  geom_point()

Basic Graphs

Use a geom to represent data points, use the geom’s aesthetic properties to represent variables. Each function returns a layer.

Line Plot

同一個主題
呈現趨勢與變化
時間序列資料

ggplot(LeBron, aes(as.numeric(YEAR), PTS)) + 
  geom_line()

可以利用geom_line、geom_ribbon等來使得視覺化結果更能傳達意思，同時使用lab()來更改主標題、x/y軸標題。

ggplot(LeBron, aes(as.numeric(YEAR), PTS)) + 
  geom_line(color='blue')+
  geom_hline(yintercept = 2000, color="red", linetype="dashed")+
  geom_ribbon(aes(ymin = 2000, ymax=PTS), fill = 'grey')+
  labs(title="LeBron James' overall Points", x='Season', y='Points')

Scatter Plot

兩個變數
呈現兩變數間的關係
正相關、負相關等等

ggplot(nba_position, aes(HEIGHT, 
                         WEIGHT, 
                         color = POSITION, 
                         alpha = 0.3,
                         shape = POSITION
                         # size = ...
                         ))+
  geom_point(position = 'jitter')

可以透過geom_point(position = 'jitter)'或是geom_jitter()來讓資料點更加分散，達到較好的呈現結果。

ggplot(sample_n(nba_position, 500), aes(HEIGHT, 
                                        WEIGHT, 
                                        color = POSITION, 
                                        alpha = 0.8, 
                                        shape = POSITION
                                        # size = ...
                                        ))+
  geom_jitter()

可以使用guides()來選擇圖例所要顯示的內容，或是是否要顯示顏色、形狀等：

ggplot(nba_position, aes(HEIGHT, 
                         WEIGHT, 
                         color = 'blue', 
                         alpha = 0.3,
                         shape = POSITION
                         # size = ...
                         ))+
  geom_point(position = 'jitter')+
  guides(
         colour = "none",
         shape = "legend", 
         alpha=FALSE
         )

Histogram

同一個主題
呈現數值分佈
表達連續性資料

使用1單位身高作為每個bin的涵蓋區間：

ggplot(nba_position, aes(x=HEIGHT)) + 
  geom_histogram(binwidth=1)

將不同位置分開，繪製histogram結合密度圖density plot以及標註平均值，可以看出不同位置之間的分布差異：

library(plyr)
mu <- ddply(nba_position, "POSITION", summarise, grp.mean=mean(HEIGHT))
ggplot(nba_position, aes(x=HEIGHT, color=POSITION)) +
geom_histogram(aes(y=..density..), position="identity", alpha=0.5, binwidth = 1)+
geom_density(alpha=0.6)+
geom_vline(data=mu, aes(xintercept=grp.mean, color=POSITION),
             linetype="dashed")

Bar Chart

一組數值資料
呈現數量分布
表達類別型資料

繪製 LeBron 歷年的比賽場數資料，並依照所屬球隊上色：

ggplot(LeBron, aes(as.numeric(YEAR), GP, fill=TEAM_ABBREVIATION)) + 
  geom_bar(stat = "identity") + 
  xlab('SEASON')

繪製 Dwyane 歷年的比賽場數資料，並依照所屬球隊上色：

ggplot(Dwyane, aes(as.numeric(YEAR), GP, fill=TEAM_ABBREVIATION)) + 
  geom_bar(position = "dodge", stat = "identity") + 
  xlab('SEASON')

繪製不同位置的球員數量長條圖：

Position_cnt <- as.data.frame(table(nba_position$POSITION))
colnames(Position_cnt)<-c("Position", "Count")
ggplot(Position_cnt, aes(Position, Count, fill=Position)) + 
  geom_bar(stat = "identity")

Box Plot

同一個主題
呈現資料的分佈
類別之間的分佈差異

ggplot(nba_position, aes(POSITION, WEIGHT, color=POSITION))+
  geom_boxplot()

使用 violin plot 也有類似效果：

ggplot(nba_position, aes(POSITION, WEIGHT, color=POSITION))+
  geom_violin(scale = 'area')

Themes

可以設置途中 line, rect, text, title, aspect.ratio, axis, legend, panel, plot, strip…etc.的樣式，舉例來說設定圖標題置中：

ggplot(LeBron, aes(as.numeric(YEAR), PTS)) + 
  geom_line(color='blue')+
  geom_hline(yintercept = 2000, color="red", linetype="dashed")+
  geom_ribbon(aes(ymin = 2000, ymax=PTS), fill = 'grey')+
  labs(title="LeBron James' overall Points", x='Season', y='Points')+
  theme(plot.title = element_text(hjust = 0.5))

或是將主題換成classic：

ggplot(LeBron, aes(as.numeric(YEAR), PTS)) + 
  geom_line(color='blue')+
  geom_hline(yintercept = 2000, color="red", linetype="dashed")+
  geom_ribbon(aes(ymin = 2000, ymax=PTS), fill = 'grey')+
  labs(title="LeBron James' overall Points", x='Season', y='Points')+
  theme_classic()

Statistics

Some plots visualize a transformation of the original data set. Use a stat to choose a common transformation to visualize. 意即繪圖前是否需要進行統計上的轉換、或是繪製統計圖：

ggplot(nba_position, aes(HEIGHT, WEIGHT))+
  geom_point(position='jitter', aes(color = POSITION))+
#  geom_smooth()
  stat_smooth(method = "auto", formula = y ~ x)

## `geom_smooth()` using method = 'gam' and formula 'y ~ s(x, bs = "cs")'

ggplot(Dwyane, aes(GP, PTS))+
  geom_point()+
  stat_quantile()

## Smoothing formula not specified. Using: y ~ x

Scales

可以去更改資料對應到aes時的對應方式，如將顏色改成連續的漸層色：

ggplot(Dwyane, aes(GP, PTS, color = PF))+
  geom_point()+
  scale_color_distiller()

或是specify要呈現在圖上的資料類別、對應的顏色、還有呈現在圖例上的名稱等：

ggplot(sample_n(nba_position, 500), aes(HEIGHT, 
                                        WEIGHT, 
                                        color = POSITION 
                                        # alpha = 0.8, 
                                        # shape = POSITION
                                        # size = ...
                                        ))+
  geom_jitter()+
  scale_color_manual(
    limits=c("Center","Forward","Guard"), # 哪些原始資料值可被視覺化？
    values=c("blue","red","green"), # 各別原始值對應的視覺呈現
    ### 定義刻度尺說明（即guide/legend）
    name="Position", # 刻度尺名稱
    breaks=c("Center","Forward","Guard"), # 哪些原始值要顯現？
    labels=c("C","F","G") # 那些原始值的標籤名稱。
  )

## Warning: Removed 38 rows containing missing values (geom_point).

Coordinates

可以更改視覺化結果的座標系統，以達到更好的視覺呈現：

ggplot(Position_cnt, aes(Position, Count, fill=Position)) + 
  geom_bar(stat = "identity")+
  coord_polar(theta = "x", direction=1 )

ggplot(Position_cnt, aes(Position, Count, fill=Position)) + 
  geom_bar(stat = "identity")+
  coord_flip()

Facets

如果想要將圖片切成子圖時，可以使用facet_grid()來處理切割的依據：

ggplot(nba_position, aes(HEIGHT, WEIGHT, color = POSITION))+
  geom_point(position='jitter')+
  facet_grid(POSITION ~ .)

DataVisualization

Yu-Hsuan Chou

2019/11/20